coredump: ensure all coredumping tasks have SIGNAL_GROUP_COREDUMP
authorOleg Nesterov <[email protected]>
Sat, 7 Nov 2015 00:32:31 +0000 (16:32 -0800)
committerLinus Torvalds <[email protected]>
Sat, 7 Nov 2015 01:50:42 +0000 (17:50 -0800)
commit5fa534c987784c4811757a34c425aff3ce3b5037
tree09aa19be1afc7e10913f29758055c2056d5d3695
parent9317bb9696566e6759203ffcaa80481b725785b3
coredump: ensure all coredumping tasks have SIGNAL_GROUP_COREDUMP

task_will_free_mem() is wrong in many ways, and in particular the
SIGNAL_GROUP_COREDUMP check is not reliable: a task can participate in the
coredumping without SIGNAL_GROUP_COREDUMP bit set.

change zap_threads() paths to always set SIGNAL_GROUP_COREDUMP even if
other CLONE_VM processes can't react to SIGKILL.  Fortunately, at least
oom-kill case if fine; it kills all tasks sharing the same mm, so it
should also kill the process which actually dumps the core.

The change in prepare_signal() is not strictly necessary, it just ensures
that the patch does not bring another subtle behavioural change.  But it
reminds us that this SIGNAL_GROUP_EXIT/COREDUMP case needs more changes.

Signed-off-by: Oleg Nesterov <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Kyle Walker <[email protected]>
Acked-by: Michal Hocko <[email protected]>
Cc: Stanislav Kozina <[email protected]>
Cc: Tetsuo Handa <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
fs/coredump.c
kernel/signal.c